$1.70 Per Worker: The Math of Not Moving

The Philippines spends about $1.70 per IT-BPM worker per year on AI transition. The size of that number is the actual bet the country is making on the risk.

The most revealing number in the Philippine AI response is $1.70.

That is the annual public commitment, per IT-BPM worker, behind the country’s flagship AI-upskilling program.

Project UNLAD — the public program meant to prepare IT-BPM workers for AI and digital-skilling needs — has a four-year budget of ₱740 million. At the BSP reference rate of roughly ₱60 to the dollar, that is about $12.3 million in total, or just over $3 million a year. Spread across a sector employing about 1.89 million people, the dedicated public commitment works out to about $1.60 to $1.70 per worker per year.1

The scale of the exposure makes that figure hard to defend. IT-BPM employs nearly 1.9 million Filipinos, contributes more than 8 percent of GDP, and generates roughly $40 billion in export revenues.2 It is one of the load-bearing pieces of the Philippine growth model.

A budget of $1.70 per worker is not denial; it is underfunding. The state has named the risk without yet financing a response.

What the $1.70 actually counts

The figure is narrow, and it needs to stay narrow. It measures dedicated public transition funding for IT-BPM workers, and it excludes private company training budgets, which sit in a different category.

IBPAP member companies reportedly spend about ₱1.4 billion a year on upskilling and learning and development, roughly $23 to $25 million depending on the exchange rate used.3 That is real money, but it is doing a different job.

Corporate L&D is built around the needs of the firm: onboarding, account delivery, compliance, productivity tools, retention, and movement into higher-value roles where the company can capture the return. That is how company training is supposed to work. A board cannot justify a training budget primarily designed to solve a national labor-market problem.

Adding the annual UNLAD allocation to the reported industry L&D spend lifts the combined figure to roughly $14 to $15 per IT-BPM worker per year. That broader training number should not be mistaken for public transition capacity.

The distinction matters because AI disruption reaches beyond firm-level skills. It is a sector-level adjustment problem: redeployment across employers, practical retraining outside current accounts, placement support, founder formation, and new domestic capability that may compete with the labor-hour model on which today’s firms still depend.

Company training improves workers inside the current structure. Public transition funding has to prepare them for the possibility that the structure itself changes. The dedicated public number for that second job is still about $1.70.

The reach problem

The coverage numbers are just as thin.

At a February Senate hearing, IBPAP said projected government-funded AI training programs for 2026 may cover approximately 68,000 learners. Against a 1.9-million-worker sector, that is about 3.6 percent of the direct workforce.4

That may be a first-year figure rather than the full ambition of the program, and it should be read generously. Even on the more generous reading, it shows how slowly the system is moving. If 68,000 learners became the annual pace, reaching half of the direct workforce would take nearly 14 years. Reaching the whole workforce would take almost 28. The likely disruption window is much shorter than that.

The risk is measured in millions of workers, in the income base of entire cities, and in the future of a $40 billion export engine. The publicly funded training reach is measured in tens of thousands.

What serious public spending looks like

Thailand is an imperfect comparison. Its economy is structured differently, and its AI policy problem is broader than Philippine IT-BPM. It is useful anyway, because it shows what a regional peer’s budget looks like when AI is treated as a serious line item.

Thailand approved a ฿25 billion AI plan for fiscal years 2026 and 2027, about $774 million over two years. The plan includes ฿6 billion for AI workforce development, ฿5 billion for AI Centers of Excellence, and ฿2 billion for a National Data Bank.5

The workforce-development line alone is the relevant comparison. At the implied exchange rate, ฿6 billion is roughly $186 million over two years, or about $93 million a year. UNLAD is about $3.1 million a year. Even before counting Thailand’s centers of excellence, data infrastructure, or broader AI program, the Thai workforce line alone is roughly 30 times larger than the Philippines’ dedicated public IT-BPM transition budget.

Again, Thailand’s number is national and UNLAD’s is sector-specific, so the comparison is imperfect. The scale gap is large enough that the imperfection does not save the Philippine position. Thailand is budgeting AI workforce development in the hundreds of millions a year; the Philippines is budgeting its most exposed services workforce in the low single-digit millions.

India is the more strategic comparison, because it faces a closer version of the same services transition. India is also exposed to AI in outsourcing, but it enters the transition with domestic champions, deeper technology firms, and a larger public capability-building program.

The IndiaAI Mission was approved with a five-year outlay of 103.7 billion rupees (10,371.92 crore), roughly $1.2 to $1.25 billion depending on exchange rate. The mission has reported 38,000 GPUs onboarded, with expansion beyond the original 10,000-GPU target. It funds compute access, datasets, foundation-model work, startup financing, safe AI initiatives, and skills programs inside the same national architecture.6

The skilling numbers need careful reading too. FutureSkills Prime has more than 2.53 million registered learners (25.3 lakh) across more than 3,000 courses and pathways.7 Registered learners are not completed transitions, and a course catalog is not deep reskilling. India’s lesson is scale and architecture, not proof that displacement disappears.

The architectural difference is what matters. India is attaching skills to compute, startups, datasets, model development, and domestic firms with the capacity to absorb and redeploy workers. That is what a transition strategy looks like when the state believes the transition is real.

What exists here is fragmented

The Philippines does have AI activity outside UNLAD, and it should be acknowledged.

DOST says it has invested ₱2.3 billion in 113 AI-related R&D projects since 2017. It has also announced plans to invest more than ₱2.6 billion in AI projects through 2028, covering healthcare, education, mobility, environment, disaster-risk reduction, industry, and emerging technology platforms.8

Those investments matter. They support research, public-sector use cases, general capacity-building, and the early shape of an AI ecosystem. DOST also says AI Pinas Summer School and SPARTA have collectively upskilled more than 49,000 Filipinos in data science and AI.9

None of this adds up to a mass transition strategy for the IT-BPM workforce. The investments are dispersed across sectors and institutions, aimed at research capacity, public applications, general AI literacy, and data-science capability. Those are necessary pieces of national AI readiness, but they are not a substitute for absorbing a concentrated labor shock in BPO.

The exposure inside IT-BPM is specific. Contact centers and routine business-process work sit directly in the path of AI substitution, augmentation, and repricing. A worker on a voice-heavy account needs more than general AI activity somewhere in the system. That worker needs a credible path into work that still pays, still uses their accumulated domain knowledge, and still keeps them inside the formal economy.

What the Philippines has, in other words, is a portfolio of projects, not a funded conversion path.

What proportionality would cost

A conservative target for what proportionality looks like: retrain half of the direct IT-BPM workforce over five years.

Half of 1.89 million workers is about 945,000 people. Spread over five years, that is roughly 189,000 workers a year through programs with enough depth to matter — instructor time, applied projects, mentorship, employer partnerships, placement support, and enough duration to turn a former agent, QA analyst, workforce planner, or operations supervisor into someone who can work higher up the AI-enabled services stack.

At $1,000 per worker, that costs about $189 million a year. At $2,000, about $378 million. At $3,000, about $567 million.10

Those numbers look large only if the underlying problem is assumed to be small. A $189 million annual program would be less than half of one percent of a $40 billion export industry. A $567 million program would be about 1.4 percent. Neither figure includes the wider urban consumption base, office demand, household credit, and local tax revenues that depend on BPO payroll.

Set against that, a public budget of roughly $3 million a year is more than modest — it is less than two percent of the lowest illustrative annual program cost above. It can fund pilots, generate credentials, and help a first cohort. It cannot plausibly absorb the labor-market risk carried by a 1.9-million-person sector.

Why industry will move too cautiously

The firms have rational reasons to move carefully.

A BPO operator has to protect margins, satisfy clients, manage attrition, and avoid retraining its own employees into jobs that may sit outside its business model. The company-level incentive is to invest enough to preserve accounts and reposition higher performers, while waiting to see how quickly clients actually reprice work around AI. That is rational at the firm level and inadequate at the national level.

The transition risk is collective. An operator that spends heavily on AI-native service design and worker redeployment while competitors wait takes the early margin hit and educates the market for everyone else. A company that trains workers for roles outside its own accounts may lose the people it just paid to prepare. A firm that builds domestic product capability competing with the labor-hour model may cannibalize current revenue before new revenue is proven.

Private firms will move at the speed of account economics. Government has to move at the speed of national exposure. That means doing more than convening stakeholders and announcing programs. It means buying down the transition risk that no individual firm has reason to carry alone.

The decision inside the number

The $1.70 figure should not be used to pretend nothing is happening. Something is happening. UNLAD exists. DOST has funded AI research. IBPAP members are spending on training. The industry knows the old model has to move.

The problem is the size of the response relative to the size of the exposure.

The Philippines built a middle class on IT-BPM payroll. That payroll supports offices, retail, apartments, transport, household credit, school fees, and local tax bases. When AI changes the labor intensity of that sector, the adjustment will move beyond call centers and shared-services floors. It will travel through the same channels the boom once moved through.

A transition budget should be sized for that income base, not only for the number of learners a pilot can process.

There are reasons someone might believe today’s spending is enough: that AI adoption will be slower than feared, that clients will keep more humans in the loop than the automation forecasts suggest, that private firms will handle most of the transition internally, or that the sector can keep growing in revenue while moving enough workers into higher-value roles to protect the domestic income base. Each of those is a bet, not a plan. The current budget is the bet the country is actually making.

At $1.70 per worker per year, that bet is that a structural transition in one of the country’s most important industries can be managed with pilot-level public spending and firm-level training incentives. The scale is disproportionate to the risk being carried.

The failure, so far, is institutional: a structural transition is being handled as an employability program. The country has the workers, the domain knowledge, and the warning signs. What it has not yet done is fund the bridge at the scale required to cross it.

That is the math of not moving.

Footnotes

  1. Philippine News Agency, “IT-BPM exec optimistic of sector’s growth amid challenges,” January 28, 2026; reports 1.89 million sector employment and a ₱740-million four-year Project UNLAD allocation. PNA — IT-BPM exec optimistic. BSP reference rate on April 21, 2026 was ₱59.950 per US dollar. BSP Reference Exchange Rate ↩

  2. Philippine News Agency, “Senate panel moves to future-proof IT-BPM sector amid AI shift,” February 18, 2026; reports around 1.9 million workers, more than 8 percent of GDP, and roughly $40 billion in export revenues. PNA — Senate panel moves to future-proof IT-BPM ↩

  3. Business Times / Bloomberg, “Amid threats, Philippines spends millions to upskill outsourcing workers,” January 28, 2026; reports about ₱1.4 billion a year in company upskilling spend and notes that each IBPAP member has a learning and development team. Business Times — Philippines spends millions to upskill outsourcing workers ↩

  4. Philstar, “IBPAP warns of job losses, contraction in IT-BPM sector,” February 25, 2026; reports IBPAP’s estimate that projected government-funded AI training programs for 2026 may cover approximately 68,000 learners against a 1.9-million-worker sector. Philstar — IBPAP warns of job losses ↩

  5. Thailand Business Information Center in Taiwan / Thai Embassy, “Thailand approves THB25bn plan to accelerate AI leadership,” September 8, 2025; summarizes the ฿25-billion two-year AI plan and its allocations, including ฿6 billion for workforce development, ฿5 billion for AI Centers of Excellence, and ฿2 billion for a National Data Bank. Thai Embassy Taiwan — Thailand approves THB25bn AI plan ↩

  6. Press Information Bureau, Government of India, “Transforming India with AI,” December 30, 2025; reports the IndiaAI Mission’s 103.7-billion-rupee five-year outlay, expressed in India as ₹10,371.92 crore, and 38,000 GPUs achieved/onboarded. PIB — Transforming India with AI ↩

  7. Press Information Bureau, Government of India, “India’s Youth Dividend in the AI Era,” February 19, 2026; reports FutureSkills Prime with more than 2.53 million registered learners, expressed in India as 25.3 lakh, and more than 3,000 courses and pathways. PIB — India’s Youth Dividend in the AI Era ↩

  8. DOST, “2025 AI Fest highlights future growth prospects for PH,” August 15, 2025; reports ₱2.3 billion invested in 113 AI-related R&D projects since 2017. DOST — 2025 AI Fest. Philippine News Agency, “DOST to invest P2.6 billion for AI projects until 2028,” June 9, 2025. PNA — DOST P2.6 billion AI projects ↩

  9. Philippine News Agency, “DOST to invest P2.6 billion for AI projects until 2028,” June 9, 2025; reports that AI Pinas Summer School and SPARTA have collectively upskilled more than 49,000 Filipinos in data science and AI. PNA — DOST P2.6 billion AI projects ↩

  10. Author scenario calculation. Half of 1.89 million workers is 945,000. Over five years, that is 189,000 workers per year. At $1,000, $2,000, and $3,000 per worker, annual costs are $189 million, $378 million, and $567 million respectively. The unit costs are illustrative, not published program estimates. ↩